https://cuicaihao.github.io/zh/tags/%E6%A2%AF%E5%BA%A6%E4%B8%8B%E9%99%8D/Recent content in 梯度下降 on C.CUI's LogHugozh-HansWed, 07 Oct 2026 07:00:00 +1100https://cuicaihao.github.io/zh/posts/2026-10-07-optimization-in-machine-learning-from-gradient-descent-to-boosting-neural-networks-and-map/Wed, 07 Oct 2026 07:00:00 +1100https://cuicaihao.github.io/zh/posts/2026-10-07-optimization-in-machine-learning-from-gradient-descent-to-boosting-neural-networks-and-map/model.fit(X, y) 隐藏了模型结构、损失、约束与求解器。本文从二维二次型出发，依次解释梯度下降、牛顿法、BFGS、正则化、Boosting、神经网络、MAP 与 Prophet，最后把同一框架延伸到大语言模型训练。